HAMSTER: visualizing microarray experiments as a set of minimum spanning trees.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression data analysis relies on visualization tools for understanding interrelationships.
• Existing methods like hierarchical clustering and k-means have limitations.
• Minimum spanning trees (MSTs) offer an alternative approach for microarray data clustering and visualization.

Purpose of the Study:

• To introduce HAMSTER (Helpful Abstraction using Minimum Spanning Trees for Expression Relations), an open-source system for generating MSTs from microarray data.
• To provide a novel visualization method that complements existing techniques like hierarchical clustering.
• To offer researchers new perspectives on gene expression data relationships.

Main Methods:

• HAMSTER recursively merges experiments to generate a set of MSTs for data visualization.
• Each MST represents a snapshot of the hierarchical clustering process.
• The system is implemented in C++ and utilizes Graphviz for MST layout.
• Scoring and ranking schemes are employed for the generated trees.

Main Results:

• HAMSTER generates visualizations that offer different insights compared to hierarchical clustering dendrograms.
• Performance analysis demonstrates the efficiency of the HAMSTER system.
• A web-based version (HAMSTER+) is available, requiring no programming knowledge.

Conclusions:

• HAMSTER can provide valuable, complementary insights in microarray analysis workflows.
• The tool is particularly useful for gradient datasets (e.g., time-series) and data showing cell/tissue relationships.
• Both the open-source C++ program and the web server variant are accessible to researchers.